Architecting Scalable ICPS for the Automotive Industry: Integrating NVIDIA AI Microservices with Eclipse Arrowhead
Eduard C. Popovici, Octavian Fratu, Alexandru Vulpe, Cosmina Stalidi, George Dan Suciu · 2025
The advancement of Industrial Cyber-Physical Systems (ICPS) in the automobile sector needs a systematic and scalable methodology for architectural design, standardization, and system interoperability. This paper proposes an ICPS architecture that incorporates NVIDIA AI Microservices, such as Triton Inference Server, AI Perception Service, and Multi-Camera Tracking (MTMC), within the Eclipse Arrowhead Framework. Using an approach based on microservices, that guarantees modular deployment, seamless data flow, and good orchestration, architecture solves significant challenges in building cars, checking their quality, and creating self-driving cars. Following ICPS engineering standards (RAMI 4.0, IIRA) our approach enhances service discovery, security, and lifecycle management. AI can thus make decisions in real time at both the edge and the cloud levels more easily. We will investigate in the future how effectively it performs, how fast it operates, and how well it interacts with other systems in car manufacturing environments. This work aims to enhance the way NVIDIA AI Microservices complement service-oriented architectures. This will create a framework that can be used for AI-driven ICPS in the car industry that is both scalable and adaptable. The architecture uses NVIDIA's AI ecosystem to automate processes, but its implementation is still a work in progress. It will be tested in real-world automotive settings in the future.